Looking at...


multi-point and point-to-point cable, copper, fibre and wireless connectivity. Software Defined Networks (SDNs) support the ability to establish private or public virtual networks and enable multiple operators and service providers to provision or acquire shared delivery and/or storage and/or server bandwidth, also known as ‘Te Cloud’. Te claim is that this reduces


delivery costs and creates additional revenue opportunities through differentiated services that can be dynamically provisioned to meet immediate dynamically changing needs. Te Software Defined Network


Layer can be conceptualised as a virtual software switch or router exercising centralised control over remote network devices. Data flow in a traditional


network is controlled by switches and routers each of which has a data plane, control plane and management plane. Te data plane, also known as the data forwarding plane, carries data packets from one port to another in accordance with a set of rules programmed into the device hardware. Te control plane applies the logic required to programme the data plane. Te management plane allows an


Software Defined Networks S


Geoff Varrall looks at the costs and benefits of Software Defined Networks


oftware Defined Networks are presently promoted as a new way of managing


administrator to reconfigure the device. A router decides on where to


route a packet or packet flow on the basis of the associated prioritisation. Tis is achieved in various ways depending on whether it is an IP flow or ATM or ethernet or commonly IP over any combination of these. An SDN controller centralises


these control and management plane functions in software running on a server. Te assumption is that the controller has a better overview of network routing functions. Tis is not very different from


traditional service provisioning in which information fed back from individual routers, for example on buffer occupancy, provides the information needed to decide on routing and traffic prioritisation. Te purpose of the abstraction


layer is therefore to make it easier for multiple parties to deliver services from common shared platforms with cost savings realised from more efficient utilisation of shared resources.


Deregulation It can be argued that software defined networks are a consequence of deregulation. Initially this involved the unbundling of long distance and local loop provision which meant that multiple level service level agreements needed to


About the author


Geoff Varrall is with the RF engineering and radio technology consultancy RTT. This article is a shortened version of RTT’s June Technology Topic. You can view the full version, and an archive of monthly topics extending back to 1998, at www.rttonline.com


34


be established and managed for the users sharing these resources or being served by these resources. Over time deregulation has


extended to almost all parts of telecommunication network provisioning including fixed and mobile wireless networks and more recently mobile broadband networks. Te principles of deregulation


are well established and understood. Deregulation increases market competition. Reduced prices stimulate market demand and improve market efficiency. Margin is maintained by improving technical efficiency. Tis is where things do not


always quite go to plan. Perversely, deregulation often


results in a decrease rather than increase in technical efficiency. Tere is for example no


convincing evidence that unbundling long distance and local loop access has realised any throughput efficiency gains. In cellular and wireless broadband, spectral allocations designed to maximise auction income have made it impossible for operators to realise theoretically achievable increases in spectral efficiency. Te result is that operating


margins get squeezed. Operators successfully shift some of the pain to the vendor supply chain but the result is industry consolidation with many countries now reverting to effective duopoly markets.


Rules and regulations In retrospect what has happened is that the costs of deregulation have generally been underestimated and the benefits of deregulation over estimated. Tis is because these are not in practice deregulated markets but regulated deregulated markets – deregulated markets where rules and regulations apply. Service level agreements provide an example of the underestimated


costs of regulated deregulation. Tese are intrinsically adversarial agreements open to dispute. Service levels need to be measured and managed, disputes need to be arbitrated and refunds need to be negotiated and agreed. And the process involves


hundreds of millions of lines of software code. We (RTT) have never come across a convincing model that captures the capital and operational cost of software code on a per million line basis. Our hunch is that the cost is probably significantly more than the savings achieved from the sharing of delivery resources and as such it should be separately identified as a liability on company balance sheets. If this is the case then the


operators in the strongest fiscal position going forward are those that own and manage their own networks. Te final technical point is that


the concept of centralised control as a mechanism for improving end-to-end delivery efficiency is not consistent with either legacy network theory or observed behavioural practice. Provided packets or packet flows


are prioritised accurately at the network edge it is perfectly valid to launch traffic into a network and allow distributed switches and router nodes to find the fastest and/ or lowest cost path. Always be wary of substituting


software for hardware. A close examination of a software


switch will yield the common sense realisation that most of the heavy lifting in these devices has to be realised with hardware accelerators. Te software functionality delivers flexibility but the associated cost will be clock cycles (power drain and latency) and expensive fast memory. Te hardware based network is


dead. Long live the hardware based network. Te traditional Telco is dead. Long live the traditional Telco.


LAND mobile August 2013


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36